Estimating carbon dioxide fluxes from temperate mountain grasslands using broad-band vegetation indices

Biogeosciences, 7, 683–694, 2010www.biogeosciences.net/7/683/2010/© Author(s) 2010. This work is distributed underthe Creative Commons Attribution 3.0 License.

Biogeosciences

Estimating carbon dioxide fluxes from temperate mountaingrasslands using broad-band vegetation indices

G. Wohlfahrt 1, S. Pilloni1, L. Hortnagl1, and A. Hammerle1,2

1Institute of Ecology, University of Innsbruck, Innsbruck, Austria2now at: ETH Zurich, Institute of Plant, Animal and Agroecosystem Sciences, Zurich, Switzerland

Received: 12 October 2009 – Published in Biogeosciences Discuss.: 27 November 2009Revised: 8 February 2010 – Accepted: 12 February 2010 – Published: 17 February 2010

Abstract. The broad-band normalised difference vegetationindex (NDVI) and the simple ratio (SR) were calculated frommeasurements of reflectance of photosynthetically active andshort-wave radiation at two temperate mountain grasslandsin Austria and related to the net ecosystem CO2 exchange(NEE) measured concurrently by means of the eddy covari-ance method. There was no significant statistical differencebetween the relationships of midday mean NEE with narrow-and broad-band NDVI and SR, measured during and cal-culated for that same time window, respectively. The skillof broad-band NDVI and SR in predicting CO2 fluxes washigher for metrics dominated by gross photosynthesis andlowest for ecosystem respiration, with NEE in between. Amethod based on a simple light response model whose pa-rameters were parameterised based on broad-band NDVI al-lowed to improve predictions of daily NEE and is suggestedto hold promise for filling gaps in the NEE time series. Rela-tionships of CO2 flux metrics with broad-band NDVI and SRhowever generally differed between the two studied grass-land sites indicting an influence of additional factors not yetaccounted for.

1 Introduction

Increases in atmospheric carbon dioxide (CO2) since the in-dustrial revolution are deemed responsible for about 60% ofobserved global warming (Solomon et al., 2007). Of the7 Pg carbon released on average each year to the atmospherethrough the burning of fossil fuels, the terrestrial biosphereabsorbs about one third (Canadell et al., 2007). Quantifyingthe spatial patterns of the net ecosystem exchange of CO2(NEE) between land ecosystems and the atmosphere and pro-

Correspondence to:G. Wohlfahrt([email protected])

jecting how NEE will be affected by likely future climateand land use is thus a critical issue in environmental science(Steffen et al., 1998; Running et al., 1999; Baldocchi et al.,2001).

Remote sensing of radiation reflected from the Earth’s sur-face from satellites allows monitoring the “status” of the en-tire planet Earth with reasonable spatial and temporal cov-erage (Malenovsky et al., 2009). Employing surface re-flectance measurements for estimating NEE, while theoret-ically very appealing (Running et al., 1999), is however notstraightforward, because NEE is the sum of two compo-nent fluxes of approximately equal magnitude: gross pri-mary productivity (GPP) and ecosystem respiration (RECO).While a clear mechanistic link can be established betweenthe reflectance (and absorption) of radiation of various wave-lengths and GPP (Gamon et al., 1992), which is drivenby the amount of absorbed quanta (Larcher, 2001), sucha link is less obvious for RECO, which dominates NEEduring nighttime (Grace et al., 2007; Malenovsky et al.,2009). For remote estimation of NEE, GPP is thus very of-ten simulated based on surface reflectance and physiologi-cal/biogeochemical modelling approaches are employed forderiving RECO (e.g. Turner et al., 2004; Olofsson et al.,2008). Validation of these models requires surface-basedmeasurements of NEE (Running et al., 1999) – these are pro-vided by the FLUXNET project (Baldocchi et al., 2001; Bal-docchi, 2008) within which more than 500 stations aroundthe globe are quantifying the NEE of a wide range of ecosys-tems by means of the eddy covariance (EC) method (Baldoc-chi et al., 1988; Aubinet et al., 2000).

A wide range of vegetation indices, calculated from re-flectance in certain wavebands, has been used for inferringGPP. Very often these make use of the strong absorption bythe photosystems (PS) I and II in the red (700 and 680 nm)waveband, such as the well-known normalised differencevegetation index (NDVI; Rouse et al., 1974) or the enhancedvegetation index (Liu et al., 1995). The more sophisticated

Published by Copernicus Publications on behalf of the European Geosciences Union.

http://creativecommons.org/licenses/by/3.0/

684 G. Wohlfahrt et al.: Remote sensing of mountain grassland carbon dioxide fluxes

photochemical reflectance index (Gamon et al., 1992) ac-counts for the fact that under certain stress conditions leavesdivert excess radiative energy from PS II to the xantophyllcycle, resulting in a change in reflectance at the 531 nm waveband (Larcher, 2001). Recently, first attempts have beenmade to relate NEE to changes in fluorescence (Moya et al.,2004; Louis et al., 2005), that is the re-emission of radiativeenergy absorbed by PS II at wavebands of 690 and 740 nm(Larcher, 2001). All these vegetation indices, except forNDVI, require reflectance to be measured with a very narrowspectral resolution of<10 nm (Grace et al., 2007). Whilemany operational hyperspectral imaging spectroradiometersare able to provide this spectral resolution, few sites in theFLUXNET project provide appropriate ground truth data.As an alternative, first probably published by Huemmrichet al. (1999), reflectance in the broad bands of photosyn-thetically active (PAR; 400–700 nm) and near-infrared (NIR;700–3000 nm) may be used to approximate the NDVI andother indices making use of the differential reflectance inthese wave band regions. Broad-band NDVI has been suc-cessfully used to track phenology (Huemmrich et al., 1999;Wang et al., 2004; Richardson et al., 2007), predict the leafarea index (Wilson and Meyers, 2007; Rocha and Shaver,2009) and GPP (Wang et al., 2004, 2009; Jenkins et al., 2007;Richardson et al., 2007). Given that most of the radiationmeasurements necessary to calculate broad-band vegetationindices, probably with the exception of reflected PAR, aremade routinely at the majority of the existing flux towers thepotential of broad-band vegetation indices has not yet beenfully exploited and, at least to our knowledge, we are notaware of published studies which used broad-band vegeta-tion indices to estimate NEE.

The overall objective of the present paper is thus to explorethe potential of broad-band vegetation indices for estimat-ing NEE and its component processes GPP and RECO. Morespecifically we aim at answering the following questions: (i)do broad-band vegetation indices exhibit a different relation-ship to CO2 fluxes as compared to narrow-band indices; (ii)how well do broad-band vegetation indices relate to NEE,GPP and RECO in comparison to commonly used predictorssuch as environmental variables or canopy characteristics;(iii) how general are the relationships between broad-bandvegetation indices and the investigated CO2 flux metrics. Tothis end we use concurrent measurements of PAR and NIRreflectance, from which we calculate broad-band NDVI andsimple ratio (SR), and NEE made at two temperate mountaingrasslands in Austria.

2 Material and methods

2.1 Site description

The study sites are located on the flat bottoms of the Stubaiand Inn valley in Western Austria close to the villages of

Neustift and Rotholz, respectively. A general characterisa-tion of the study sites is given in Table 1. Neustift is cutthree times a year with occasional light grazing in Octo-ber, while Rotholz, due to the lower elevation and longervegetation period, is cut three times a year with significantgrazing in autumn (mid September to mid October). Bothsites receive only (solid and liquid) organic manure, usuallyin late autumn. Plant species composition is fairly similar,most of the above-ground biomass being made up by around10 dominant forb and grass species (Wohlfahrt et al., 2008).However, Rotholz is dominated by grasses (40–60%), whilegrasses make up only 20–40% at Neustift.

2.2 Net ecosystem CO2 exchange (NEE)

Continuous eddy covariance (EC) measurements of the NEEhave been made at Neustift and Rotholz since March 2001and December 2007, respectively, and continue as of thiswriting. Within the frame of this study we used data from2006–2008 (Neustift) and 2008 (Rotholz), as the correspond-ing multispectral reflectance (Neustift 2006) and radiationmeasurements (Neustift 2007–2008, Rotholz 2008) wereavailable during these periods. CO2 fluxes were measuredusing the EC method (Baldocchi et al., 1988) following theprocedures of the EUROFLUX project (Aubinet et al., 2000).For further details regarding instrumentation, data treatmentand quality control we refer to our previous publications(Wohlfahrt et al., 2008; Haslwanter et al., 2009). Negativefluxes represent transport from the atmosphere towards thesurface, positive ones the reverse.

Half-hourly NEE measurements were used to derive threegroups of CO2 flux metrics:

1. For consistency with the time-window used for calculat-ing broad-band vegetation indices (see below) we calcu-lated midday mean NEE (NEEmdm) for the time period10:00–14:00 Central European Time (CET);

2. Daily sums of NEE (NEEd) were derived by usinga standardised approach for filling gaps (Falge et al.,2001; Moffat et al., 2007) as described in Wohlfahrtet al. (2008); ecosystem respiration (RECO) was cal-culated from nighttime NEE by extrapolation to day-time temperatures (Reichstein et al., 2005) and then in-tegrated over each day (RECOd); gross primary produc-tivity (GPP) was calculated as NEE-RECO and then in-tegrated to daily values (GPPd); further details of thisprocedure are given in Wohlfahrt et al., 2008.

3. Light-response curve analysis is a powerful tool forcharacterising key ecosystem functional properties(Ruimy et al., 1995). In order to relate NEE to PARwe fitted (using the Levenberg-Marquard algorithm) thedata (pooled into five-day blocks of data) to the follow-ing sigmoid model (Smith, 1937):

Biogeosciences, 7, 683–694, 2010 www.biogeosciences.net/7/683/2010/

G. Wohlfahrt et al.: Remote sensing of mountain grassland carbon dioxide fluxes 685

Table 1. General characterisation of the two study sites.

Neustift Rotholz

Latitude 47◦07′N 47◦23′NLongitude 11◦19′E 11◦48′EElevation (m a.s.l.) 970 523Mean annual temperature (◦C) 6.5 8.2Mean annual precipitation (mm) 852 1151Vegetation type Pastinaco-Arrhenatheretum Lolietum multifloraeSoil type (FAO classification) Gleyic Fluvisol Gleyic Fluvisol

NEE=αQPARFsat√

F 2sat+(αQPAR)2

+Reco (1)

Here α represents the apparent quantum yield(mol CO2 mol photons−1), Fsat the asymptotic value ofthe gross primary production (GPP) at high irradiance(µmol CO2 m−2 s−1; referred to as GPPmax in the follow-ing), QPAR the PAR (µmol photons m−2 s−1) and Reco theecosystem respiration (µmol CO2 m−2 s−1). Equation (1)represents an unbiased approximation to the light responseof a wide range of ecosystems and is much better suited thanthe often used rectangular and non-rectangular hyperbolicmodels (A. Moffat, personal communication, 2009).

2.3 Multispectral reflectance and narrow-bandvegetation indices

Multispectral reflectance was measured with a portable spec-troradiometer (FieldSpec Handheld, ASD Inc., Boulder, CO,USA) with a 3 nm spectral resolution (325–1075 nm range).Measurements of nadir vegetation radiance and the zenithsky irradiance, from which reflectance was calculated, weremade with a cosine diffuser foreoptic (170◦ field of view)at the same representative vegetation spot around noon on20 clear days between April and October 2006. To this endthe spectroradiometer was mounted 1.5 m above ground on ahorizontally protruding rotating arm on a tripod allowing al-ternating downward and upward measurements without dis-turbing the vegetation and interfering with the reflectancemeasurement (Gianelle et al., 2009).

For each measurement date the normalised difference veg-etation index (NDVI) and the simple ratio (SR) were calcu-lated based on five replicate measurements (coefficients ofvariation generally<5%) of the reflectance in the NIR (770–780 nm) and red (670–680 nm) wavebands (rnir, rred):

NDVI =(rnir −rred)

(rnir +rred)(2)

SR=rnir

rred(3)

2.4 Radiation measurements and broad-bandvegetation indices

Incoming total and diffuse photosynthetically active radia-tion (PAR) were measured by photodiodes (BF2H, Delta-TDevices Ltd, Cambridge, UK) mounted at 2 m above ground.PAR incident on the ground surface (PARsoil) and reflectedPAR (PARrefl) were measured by a line quantum sensor (10individual sensors providing an averaged output) and a singlequantum sensor (LQS7010Sun and QSOSun, respectively,Apogee Instruments, Logan, USA). The line quantum sensormeasuring PARsoil was mounted in a U-shaped aluminiumprofile, which was placed about 0.01 m below the soil surfacein order to measure the PAR incident on the soil surface. Thepyranometers of the CNR-1 net radiometer (CNR-1, Kippand Zonen, Delft, The Netherlands) were used for measur-ing the up- and down-welling global radiation (Rg, 305 to2800 nm). All radiation sensors were inter-calibrated prior tothe experiment in order to avoid systematic biases.

PAR photon flux densities (µmol photons m−2 s−1) wereconverted to energy units (J m−2 s−1) using 4.55 µmol J−1

(Goudriaan and Van Laar, 1994) for both incident and re-flected PAR, neglecting any spectral differences between thetwo. Ignoring any radiation in the 305–400 nm range, nearinfrared radiation (NIR) was then calculated fromRg by sub-tracting the PAR component:

NIRin = Rgin −PARin (4)

NIRrefl = Rgrefl −PARrefl (5)

Broad-band NDVI and SR were then calculated by replac-ing rnir and rred in Eqs. (2) and (3) with NIRrefl/NIRin andPARrefl/PARin, respectively. In order to minimise the ef-fect of daily and seasonal changes in the sun’s elevation,the broad-band NDVI and SR were averaged around middaybetween 10:00–14:00 CET. In addition, data were excludedwhen the site was covered by snow or when it had rainedfour hours prior or during the midday averaging period.

The calculation of broad-band vegetation indices as de-scribed above involves some simplifications/compromises –

www.biogeosciences.net/7/683/2010/ Biogeosciences, 7, 683–694, 2010


Fig. 1. Seasonal variation of(a, e) reflectance in photosynthetically active (PAR, open symbols) and near-infrared radiation (NIR, closedsymbols),(b, f) normalized difference vegetation index (NDVI),(c, g) simple ratio (SR), and(d, h) fraction of absorbed PAR (fPAR) andmeasured (symbols) and simulated (lines) GAI. Snow cover periods (d, h) and cutting/grazing events are indicated by grey horizontal andvertical bars, respectively.

in particular neglecting spectral differences in up- and down-welling PAR, the conversion of PAR from photon to energyflux density and lumping the 305–400 nm waveband to theNIR component has the potential to systematically bias re-sults. Because the precise magnitude of these uncertainties ishard to quantify and because we also wanted to account foradditional potential sources of uncertainty (e.g. due to shiftsin sensor sensitivity with time) we have conducted an uncer-tainty analysis by arbitrarily assuming a 20% uncertainty inthe NIR and PAR reflectance, either separately or in combi-nation with uncertainties of opposing magnitude. To this endreflectance values typical for small (rNIR:0.21, rPAR:0.08)and large (rNIR:0.45, rPAR:0.04) amounts of above-groundphytomass (Fig. 1) were used. A 20% uncertainty in eitherthe NIR or PAR reflectance resulted in uncertainties of up to20 and 25% for NDVI and SR, respectively. For NDVI thisuncertainty was smaller for large NDVI values (only 4%).Uncertainties up to 38 and 50%, for NDVI and SR respec-tively, were observed when uncertainties in NIR or PAR re-flectance were assumed to occur jointly and in opposite di-rections.

The fraction of absorbed photosynthetically active radia-tion (fPAR) was calculated by assuming a soil reflectance of15% (Goudriaan, 1977) and again averaged around middaybetween 10:00–14:00 CET.

fPAR=PARin −PARrefl−0.85PARsoil

PARin(6)

The amount of absorbed PAR (aPAR) was calculated fromfPAR by multiplication with incident PAR.

2.5 Ancillary data

Further supporting meteorological measurements includedair temperature (TA) and humidity (RH) at 2 m above groundand soil temperature (TS) at 0.05 m depth, measured by themeans of a combined temperature/humidity sensor (RFT-2,UMS, Munich, Germany) and an averaging soil thermocou-ple (TCAV, Campbell Scientific, Logan, UT, USA), respec-tively. Further, soil water content (SWC; at 0.1 m soil depth)(ML2x, Delta-T Devices, Cambridge, UK) and precipitation(Precip) (52202, R. M. Young, Traverse City, MI, USA) weremeasured.

Green, photosynthetically active stems make up a consid-erable fraction of the above-ground biomass (Wohlfahrt etal., 2001). Therefore we use the green area index (GAI,m2 m−2), which comprises both leaves and green stems, in-stead of the commonly used leaf area index for quantifyingthe amount of photosynthetically active plant surface. TheGAI was assessed (i) in a destructive fashion by clipping ofsquare plots of 0.09 m2 (3–5 replicates) and subsequent plantarea determination (Li-3100, Li-Cor, Lincoln, NE, USA) and(ii) from measurements of canopy height which was relatedto destructively measured GAI. Continuous time series of theGAI were derived by fitting appropriate empirical functions,separately for each growing phase, to measured data.



3 Results

Seasonal canopy development, governed by the cuts and thesubsequent plant regrowth, of the investigated grasslands ledto characteristic patterns in reflectance and vegetation in-dices (Fig. 1). Representative examples of correspondingchanges in multispectral reflectance are pictured in Fig. 2,which shows that when little vegetation cover was present acomparably larger fraction of the radiation in the PAR wasreflected, while reflectance was smaller in the NIR result-ing in low narrow-band NDVI and SR values (compare toEqs.2 and3). As the amount of above-ground plant matterincreased, reflectance decreased rapidly in the PAR and in-creased in the NIR (Fig. 2), causing narrow-band NDVI andSR to increase.

Broad-band reflectance and reflectance indices showeda qualitatively similar behaviour: Mean midday NIR re-flectance increased asymptotically with increasing GAI andfPAR towards the cutting events, ranging between 0.15–0.40 and 0.20–0.45 at Neustift and Rotholz, respectively(Fig. 1). In contrast, PAR reflectance decreased asymptoti-cally from around 0.10 down to 0.04 during the same peri-ods (Fig. 1). The cutting events reduced NIR and increasedPAR reflectance, setting values back close to springtime lev-els. Mirroring the trends in NIR and PAR reflectance, boththe NDVI and SR increased asymptotically during the vari-ous growing phases. NDVI started at approximately 0.45 andlevelled off at around 0.85, SR increased from values around2 up to 13 (Fig. 1). Note that reflectance values and vegeta-tion indices during the phase of intensive grazing at Rotholzremained elevated because we failed to adequately simulategrazing below the (fenced) radiation sensors (Fig. 1). In anattempt to rectify this situation we cut the grass down to theactual grazing level in early October, resulting in an imme-diate decrease in reflectance values and vegetation indices(Fig. 1). Because of the mismatch with NEE, representativeof the grazed area, data after the beginning of grazing wereexcluded from the following analysis.

Midday means of NEE (averaged between 10:00–14:00 CET) showed a sharp increase in carbon uptake inearly spring and after the cutting events, which levelled offat about−26 µmol CO2 m−2 s−1 at both sites (Fig. 3). Thecutting events strongly reduced midday mean NEE for sev-eral days or even caused it to change sign (maximum lossaround 12 µmol CO2 m−2 s−1). After the third cut both sitesagain turned into sinks over the midday hours, but did notreach CO2 uptake values comparable to those in the previ-ous phases (Fig. 3). These were further reduced at Rotholzduring a period of intensive grazing. Qualitatively similarpatterns emerged from the analysis of daily NEE (Fig. 3),except for that maximum daily NEE tended to decrease overthe course of the vegetation period at Rotholz. This wascaused by RECO progressively increasing until the begin-ning of September, while daily GPP largely followed the sea-sonal course of the GAI (Fig. 1).

Fig. 2. Representative examples of multispectral reflectance for arange of GAI (green area index) values. Data have been measuredat the study site Neustift during June–July 2006. The correspond-ing NDVI(SR) values are (in order of ascending GAI): 0.56(3.52),0.67(5.13), 0.70(5.76), 0.76(7.31).

As shown in Fig. 4, the relationships between narrow- andbroad-band NDVI and SR, which were measured and aver-aged around noon respectively, and midday mean NEE werevery similar (data available only at Neustift). This is con-firmed by a comparison of the slopes andy-intercepts oflinear regressions between NDVI and SR against NEEmdm,which were statistically indistinguishable (p >0.63) betweenboth methods.

A linear regression analysis shows that broad-band veg-etation indices (i.e. NDVI and/or SR) averaged over middayhours explained most of the variability in NEE averaged overthat same period (NEEmdm) and in daily NEE, followed byfPAR and aPAR, GAI, and finally by the investigated envi-ronmental parameters (Table 2). In contrast, while broad-band NDVI and SR were still excellent predictors for dailyGPP (the best for Rotholz), the difference to the environmen-tal variables was lower. This was even more so the case fordaily RECO, where most of the variability was explained byair and soil temperatures, and broad-band NDVI and SR ex-plained hardly as much or even less of the variability as com-pared to day length (Table 2). A similar result was obtainedfor the ecosystem respiration parameter in the light responsecurve model (Reco, Eq. 1). The other parameters of Eq. (1),the apparent quantum yield (α) and the GPP at saturatinglight intensity (GPPmax) were generally well explained bythe broad-band vegetation indices (Table 1). The slopes andy-intercepts of these linear relationships (Figs. 4, 5 and 6)were significantly different between Neustift und Rotholz forall dependent variables in Table 2 except for GPPd (slope),NEEmdm (y-intercept),α and Reco (slope andy-intercept).A step-wise multiple regression (data not shown) yieldedhardly any improvement for ecosystem respiration (RECOdandReco), while the largest improvement in explained vari-ance (up to 33%) was observed for daily NEE.



Fig. 3. Seasonal variation of(a, d)midday mean net ecosystem CO2 exchange (NEEmdm), (b, e)daily net ecosystem CO2 exchange (NEEd),and(c, f) daily gross primary productivity (GPPd, closed symbols) and daily ecosystem respiration (RECOd, open symbols). Snow coverperiods (c, f) and cutting/grazing events are indicated by grey horizontal and vertical bars, respectively.

Table 2. Results (Pearson correlation coefficients) of linear regression analysis.

Neustift Rotholz

NEEmdm NEEd GPPd RECOd α GPPmax Reco NEEmdm NEEd GPPd RECOd α GPPmax Reco

NDVI −0.79 −0.66 −0.78 0.51 0.68 −0.79 0.41 −0.90 −0.68 −0.95 0.83 0.77 −0.96 0.81SR −0.76 −0.69 −0.77 0.45 0.64 −0.78 0.36 −0.90 −0.72 −0.93 0.76 0.70 −0.96 0.75GAI −0.66 −0.63 −0.84 0.65 0.64 −0.83 0.60 −0.77 −0.66 −0.91 0.79 0.70 −0.90 0.77fPAR −0.79 −0.72 −0.72 0.34 0.57 −0.70 0.21 −0.88 −0.70 −0.91 0.75 0.69 −0.95 0.73aPAR −0.67 −0.68 −0.79 0.50 0.34 −0.86 0.61 −0.84 −0.70 −0.91 0.75 0.61 −0.95 0.79PAR −0.31 −0.44 −0.60 0.48 0.13 −0.68 0.69 −0.57 −0.50 −0.78 0.72 0.59 −0.87 0.84TA −0.22 −0.21 −0.63 0.82 0.49 −0.64 0.89 −0.35 −0.24 −0.77 0.92 0.75 −0.76 0.98TS −0.18 −0.18 −0.64 0.87 0.51 −0.59 0.87 −0.47 −0.27 −0.82 0.97 0.80 −0.78 0.98VPD −0.20 −0.30 −0.50 0.49 0.25 −0.65 0.70 −0.26 −0.26 −0.57 0.62 0.49 −0.69 0.81SWC 0.32 0.31 0.49 −0.45 −0.11 0.38 −0.31 −0.15 −0.18 −0.37 0.39 0.36 −0.22 0.22DL n.a. −0.35 −0.70 0.77 n.a. n.a. n.a. n.a. −0.44 −0.80 0.81 n.a. n.a. n.a.n 167 346 346 346 89 89 89 86 149 149 149 43 43 43

Bold numbers indicate that regressions are statistically significant (p <0.05); NEEmdm . . . midday mean net ecosystem CO2 exchange,NEEd . . . daily net ecosystem CO2 exchange, GPPd . . . daily gross primary productivity, RECOd . . . daily ecosystem respiration,α

. . . apparent quantum yield, GPPmax . . . gross primary productivity at saturating light intensity,Reco . . . ecosystem respiration, NDVI

. . . broad-band normalised difference vegetation index, SR . . . broad-band simple ratio, GAI . . . green area index, fPAR . . . fraction ofabsorbed photosynthetically active radiation, aPAR . . . absorbed photosynthetically active radiation, PAR . . . incident photosyntheticallyactive radiation, TA . . . air temperature, TS . . . soil temperature, VPD . . . vapour pressure deficit, SWC . . . soil water content, DL . . . daylength,n . . . number of samples.

Because broad-band vegetation indices explained only be-tween 44–52% of the variability in daily NEE at both sites,Eq. (1) and the best-fitting relationships of the parametersof Eq. (1) in Table 2, which are shown in Fig. 6, were usedto predict daily NEE. Daily NEE predicted with this modelexplained 65 and 80% of the variability in measured daily

NEE for Neustift and Rotholz, respectively (Fig. 6; compareto results shown in Table 2). The slopes andy-interceptsof linear regressions between simulated and measured dailyNEE were not significantly different (p <0.05) from unityand zero, respectively, for Rotholz, but for Neustift.



4 Discussion

Concurrent measurements of PAR and NIR reflectance, fromwhich broad-band NDVI and SR were calculated, and NEEwere made at two temperate mountain grasslands in Austria.The objective of this study was to explore the potential ofbroad-band NDVI and SR for predicting NEE and its twocomponent processes, GPP and RECO. Our main findingswere: i) there was no significant statistical difference be-tween the relationship of midday mean NEE to narrow- andbroad-band NDVI and SR, measured during and calculatedfor that same time window, respectively; ii) the skill of broad-band NDVI and SR in predicting CO2 fluxes was higher formetrics dominated by gross photosynthesis and lowest forecosystem respiration, with NEE in between; iii) relation-ships of CO2 flux metrics with broad-band NDVI and SRgenerally differed between the two studied grassland sites.

The utility of broad-band as opposed to narrow-band veg-etation indices depends on how well the former comparewith the latter (and how sensitive the process to be in-ferred is to these vegetation indices). While several studieshave compared broad-band vegetation indices against theirnarrow-band counterparts based on satellite data (Wang etal., 2004; Jenkins et al., 2007; Rocha and Shaver, 2009),there are actually only two studies we are aware of that com-pared in-situ (i.e. without the confounding effects of atmo-spheric corrections) measured broad- with narrow-band veg-etation indices: Both Huemmrich et al. (1999) and Rochaand Shaver (2009) reported acceptable correspondence be-tween broad- and narrow-band NDVI measured using radi-ation sensors and a spectroradiometer above several borealforest ecosystems and along a tundra burn severity gradientin Alaska, respectively. Our study, while not able to directlycompare broad- and narrow-band vegetation indices as thesehave been measured during different years, lends further sup-port to the utility of broad-band vegetation indices as wewere able to show that the relationship to midday mean NEEdid not significantly differ from the one determined frommultispectral (narrow-band) data (Fig. 4). Reflectance in thebroad spectral regions of PAR and NIR are thus obviouslyable to capture many of the characteristic features of narrow-band vegetation indices despite integrating over much largerwaveband region (Huemmrich et al., 2009).

NDVI and SR have been successfully related to gross pho-tosynthesis (e.g. Gianelle et al., 2009) because they math-ematically contrast the strong absorption by the photosys-tems I and II in the red which drives photosynthesis (Larcher,2001), with the large reflectance in the NIR, which is largelythe result of photon scattering within the leaf tissue at the air-cell interfaces of the mesophyll (Asner, 1998). It thus comesas no huge surprise that gross photosynthesis at saturatinglight intensity (GPPmax) and daily GPP, but also middaymean NEE, which bears a considerable “assimilatory signa-ture”, were highly correlated with broad-band NDVI and/orSR (Table 2). Conversely, vegetation indices were poor pre-

Fig. 4. Relationship between normalized difference vegetation in-dex (NDVI, upper panel) and simple ratio (SR, lower panel) andmidday mean net ecosystem CO2 exchange (NEEmdm).

dictors of ecosystem respiration, because reflectance is notdirectly related to the various, both auto- and heterotrophic,component processes which contribute to ecosystem respira-tion (Grace et al., 2007). Seemingly, the now well establishedlink between ecosystem respiration and photosynthesis viathe supply of fresh substrate for assimilation (e.g. Hogberg etal., 2001; Bahn et al., 2009) was not exerting as strong a con-trol as compared to temperature (Table 2), which affects thespeed of metabolic reactions (Davidson and Janssens, 2006).Soil water content, affecting the accessibility of organic sub-strates in the soil (Davidson and Janssens, 2006), did not playa major role at our sites, in accordance with the findings ofWohlfahrt et al. (2008) for Neustift.

For deriving and monitoring yearly and longer-term NEEby remote sensing it would be desirable to predict daily NEE(Running et al., 1999). Very often, however, vegetation in-dices are poorly related to daily NEE (44–52% explainedvariance in the present case; Table 2), which reflects thegood and poor correlations with GPP and RECO, respec-tively (Olofsson et al., 2008). Daily NEE, furthermore, suf-fers from an unwanted influence of the length of the day-light period, although it was not as large as for daily GPP



Fig. 5. Relationship between(a–c)normalized difference vegetation index (NDVI) and(d–f) simple ratio (SR, lower panel) and (a, d) dailynet ecosystem CO2 exchange (NEEd), (b, e) daily gross primary productivity (GPPd), and (c, f) daily ecosystem respiration (RECOd).Closed and open symbols represent the study sites Neustift and Rotholz, respectively.

Fig. 6. Relationship between(a, b) normalized difference vegetation index (NDVI) and the apparent quantum yield (α) and gross primaryproductivity under saturating light conditions (GPPmax), (c) air temperature and ecosystem respiration (Reco), and (d) correspondencebetween the measured and simulated (using Eq. 1 and relationships shown in panels a–c) daily net ecosystem CO2 exchange.



Fig. 7. Example illustrating the potential of Eq. (1) and the relationships shown in Fig. 6 for gap-filling of NEE time series. Data are from2008 for the study site Rotholz and show(a) the half-hourly measured NEE (symbols) and the midday average broad-band NDVI (solidline), (b) daily average PAR (open symbols) and air temperature (closed symbols),(c) daily NEE simulated based on various gap-fillingapproaches and(d) daily and 28-day-integrated (inset) NEE gap-filled with various approaches. MDV refers to the mean diurnal variationmethod and LT to light- and temperature-dependent regression models (Falge et al., 2001). The vertical dotted lines in panels (a) and (c)indicate the two-week blocks of data used for the MDV and LT approaches. See text for further details.

and RECO, where the length of the daylight period alreadyexplained>50% of the variability in data (Table 2). The cor-relation with daily NEE, as with all daily metrics, may alsobe lower because vegetation indices calculated from middaymean reflectance values may not well reflect CO2 exchangeduring other times of the day (Sims et al., 2005). Accord-ingly, daily NEE was the metric which profited most fromincluding additional predictors in a multiple linear regres-sion (r2 increased by up to 33%). Besides this statistical ap-proach, predictions of daily NEE could also be improved,resulting in 64–80% explained variance, by using a moremechanistic light response curve model (Eq. 1) with its pa-rameters modelled (Fig. 6) as a function of broad-band NDVI(α, GPPmax) and air temperature (Reco). However, in orderto drive this model, sub-daily (i.e. half-hourly) PAR is re-quired as input, which may limit its applicability. In fact,this approach may hold more promise for reducing the uncer-tainty of gap-filled data products (Falge et al., 2001; Moffatet al., 2007). In particular for long data gaps, which repre-sent a large source of uncertainty (Richardson and Hollinger,2007), or when rapid changes in ecosystem CO2 exchangeoccur, such as with the investigated grasslands (Wohlfahrt etal., 2008) or during leaf-out of deciduous forests (Richard-son et al., 2007), broad-band vegetation indices may providea good proxy for changes in ecosystem physiological activity(Wang et al., 2009) and could be assimilated into the gap-

filling procedure. An example for this is given in Fig. 7,which shows (using data from Rotholz) differences in gap-filled NEE for a four-week period following the third cut in2008 between the proposed approach (i.e. Eq. 1 and the re-lationships shown in Fig. 6) and the mean diurnal variation(MDV) and combined light/temperature regression (LT) ap-proach (Falge et al., 2001). For the latter two approaches, afrequently used fixed two-week time window was used. Asopposed to the NDVI-based approach, which used daily mid-day average NDVI as input, this caused an overestimationand underestimation of CO2 uptake at the beginning and endof the two-week blocks of data, respectively (Fig. 7c). Theconsequences of these differences for gap-filling depend onthe number, length and distribution of gaps. In the presentexample, with around 40% gaps predominately during night-time, NEE was similar for the MDV and LT approach and25% lower for the NDVI-based approach (Fig. 7d).

The relationships between the various CO2 flux met-rics and broad-band NDVI and SR were generally site-specific (Figs. 4, 5 and 6), making the application of theserelationships to other grassland ecosystems problematic.This is not uncommon for purely statistical relationships(e.g. McMichael et al., 1999) and very much likely re-sults from unaccounted differences between the two sitesin terms of ecosystem structure, e.g. species composition inconjunction with species-specific differences in leaf optical



properties or leaf angles, and function, e.g. the relationshipbetween leaf biophysical/biochemical properties and pho-tosynthesis (Asner, 1998). Because of the relatively largeamounts of above-ground phytomass we do not believe thatthe broad-band Enhanced Vegetation Index, recently put for-ward by Rocha and Shaver (2009) because of its ability toaccount for differences in soil background reflectance, wouldbe able to remedy these differences. Rather, a more mecha-nistic treatment of the relationships between reflectance invarious wavebands and NEE and its component processes,as done in some process-oriented models (e.g. Van der Tol etal., 2009), will be required to this end.

Given that most of the radiation measurements necessaryto calculate broad-band vegetation indices, probably with theexception of reflected PAR, are made routinely at the major-ity of the existing FLUXNET sites it would require relativelylittle extra effort to extend these measurements to a largernumber of sites. As our uncertainty analysis (see Sect. 2.4)showed that both vegetation indices, but in particular SR,were very sensitive to errors in PAR and NIR reflectance (inparticular if both change in opposing directions), any imple-mentation of such measurements should be accompanied byan appropriate (network-wide) quality control protocol.

Acknowledgements.This study was financially supported by theAustrian National Science Fund under contracts P17560 andP19849 and the Tyrolean Science Fund under contracts Uni-404/33and Uni-404/486. A. Hammerle acknowledges financial supportthrough a DOC fellowship by the Austrian Academy of Sciencesand a post-graduate fellowship by the University of Innsbruck;L. Hortnagl acknowledges financial support through a PhD fellow-ship by the University of Innsbruck. D. Gianelle and L. Vescovoare thanked for loaning of the portable spectroradiometer. FamilyHofer (Neustift, Austria) and the Landeslehranstalt Rotholz,particularly H. Haas, are acknowledged for granting us access tothe study sites. Long-term temperature and precipitation data havebeen generously provided by the Austrian Hydrographic Service.

Edited by: A. Arneth

References

Asner, G. P.: Biophysical and biochemical sources of variability incanopy reflectance, Remote Sens. Environ., 64, 234–253, 1998.

Aubinet, M., Grelle, A., Ibrom, A., Rannik,U., Moncrieff, J., Fo-ken, T., Kowalski, A. S., Martin, P. H., Berbigier, P., Bernhofer,Ch., Clement, R., Elbers, J., Granier, A., Grunwald, T., Morgen-stern, K., Pilegaard, K., Rebmann, C., Snijders, W., Velentini,R., and Vesala, T.: Estimates of the annual net carbon and waterexchange of forests: The EUROFLUX methodology, Adv. Ecol.Res., 30, 113–175, 2000.

Bahn, M., Schmitt, M., Siegwolf, R., Richter, A., and Bruggemann,N.: Does photosynthesis affect grassland soil-respired CO2 andits carbon isotope composition on a diurnal timescale?, New Phy-tologist, 182, 451–460, 2009.

Baldocchi, D. D.:“Breathing” of the terrestrial biosphere: Lessonslearned from a global network of carbon dioxide flux measure-ment systems, Aust. J. Bot., 56, 1–26, 2008.

Baldocchi, D. D., Falge, E., Gu, L., Olson, R., Hollinger, D.,Running, S., Anthoni, P., Bernhofer, C., Davis, K., Fuentes, J.,Goldstein, A., Katul, G., Law, B., Lee, X., Malhi, Y., Meyers,T., Munger, J. W., Oechel, W., Pilegaard, K., Schmid, H. P.,Valentini, R., Verma, S., Vesala, T., Wilson, K., and Wofsy, S.:FLUXNET: a new tool to study the temporal and spatial variabil-ity of ecosystem-scale carbon dioxide, water vapor and energyflux densities, B. Am. Meteorol. Soc., 82, 2415–2435, 2001.

Baldocchi, D. D., Hicks, B. B., and Meyers, T. P.: Measuringbiosphere-atmosphere exchanges of biologically related gaseswith micrometeorological methods, Ecology, 69, 1331–1340,1988.

Canadell, J. G., Pataki, D. E., Gifford, R., Houghton, R. A., Luo,Y., Raupach, M. R., Smith, P., and Steffen, W.: Saturation of theterrestrial carbon sink, in: Terrestrial Ecosystems in a ChangingWorld, edited by: Canadell, J. G., Pataki, D. E., and Pitelka L. F.,Springer, Germany, 59–78, 2007.

Davidson, E. A. and Janssens, I. A.: Temperature sensitivity of soilcarbon decomposition and feedbacks to climate change, Nature,440, 165–173, 2006.

Falge, E., Baldocchi, D., Olson, R., Anthoni, P., Aubinet, M., Bern-hofer, Ch., Burba, G., Ceulemans, R., Clement, R., Dolman, H.,Granier, A., Gross, P., Grunwald, T., Hollinger, D., Jenson, N.-O., Katul, G., Keronen, P., Kowalski, A., Lai, C. T., Law, B.,Meyers, T., Moncrieff, J., Moors, E. J., Munger, W., Pilegaard,K., Rannik,U ., Rebmann, C., Sukyer, A., Tenhunen, J., Tu, K.,Verma, S., Vesala, T., Wilson, K., and Wofsy, S.: Gap fillingstrategies for defensible annual sums of net ecosystem exchange,Agr. Forest Meteorol., 107, 43–69, 2001.

Gamon, J. A., Penuelas, J., and Field, C. B.: A narrow-wavebandspectral index that tracks diurnal changes in photosynthetic effi-ciency, Remote Sens. Environ., 41, 35–44, 1992.

Gianelle, D., Vescovo, L., Marcolla, B., Manca, G., and Cescatti,A.: Ecosystem carbon fluxes and canopy spectral reflectance ofa mountain meadow, Int. J. Remote Sens., 30, 435–449, 2009.

Goudriaan, J.: Crop Micrometeorology: A Simulation Study. Wa-geningen, Centre for Agricultural Publishing and Documenta-tion, 1977.

Goudriaan, J. and van Laar, H. H.: Modelling potential crop growthprocesses, Kluwer Academic Publishers, Dordrecht, 1994.

Grace, J., Nichol, C., Disney, M., Lewis, P., Quaife, T., andBowyer, P.: Can we measure terrestrial photosynthesis fromspace directly, using spectral reflectance and fluorescence?,Global Change Biol., 13, 1484–1487, 2007.

Hammerle, A., Haslwanter, A., Tappeiner, U., Cernusca, A., andWohlfahrt, G.: Leaf area controls on energy partitioning of a tem-perate mountain grassland, Biogeosciences, 5, 421–431, 2008,http://www.biogeosciences.net/5/421/2008/.

Haslwanter, A., Hammerle, A., and Wohlfahrt G.: Open- vs. closed-path eddy covariance measurements of the net ecosystem carbondioxide and water vapour exchange: a long-term perspective,Agr. Forest Meteorol., 149, 291–302, 2009.

Hogberg, P., Nordgren, A., Buchmann, N., Taylor, A. F., Ekblad,A., Hogberg, M. N., Nyberg, G., Ottoson-Lofvenius, M., andRead, D. J.: Large-scale forest girdling shows that current pho-tosynthesis drives soil respiration, Nature, 411, 789–792, 2001.


http://www.biogeosciences.net/5/421/2008/


Huemmrich, K. F., Black, T. A., Jarvis, P. G., McCaughey, J. H.,and Hall, F. G.: High temporal resolution NDVI phenologyfrom micrometeorological radiation sensors, J. Geophys. Res.,104(D22), 27935–27944, 1999.

Jenkins, J. P., Richardson, A. D., Braswell, B. H., Ollinger, S.V., Hollinger, D. Y., and Smith, M.-L.: Refining light-use effi-ciency calculations for a deciduous forest canopy using simul-taneous tower-based carbon flux and radiometric measurements,Agr. Forest Meteorol., 143, 64–79, 2007.

Larcher, W.:Okophysiologie der Pflanzen, Eugen Ulmer, Stuttgart,2001.

Liu, H. Q. and Huete, A. R.: A feedback based modification of theNDVI to minimize canopy background and atmospheric noise,IEEE T. Geosci Remote, 33, 457–465, 1995.

Louis, J., Ounis, A., Ducret, J. -M. Evain, S., Laurila, T., Thum,T., Aurela, M., Wingsle, G., Alonso, L., Pedros, R., and Moya,I.: Remote sensing of sunlight induced chlorophyll fluorescenceand reflectance of Scots pine in the boreal forest during springrecovery, Remote Sens. Environ., 96, 37–48, 2005.

McMichael, C. E., Hope, A. S., Stow, D. A., Fleming, J. B., Vourli-tis, G., and Oechel, W. C.: Estimating CO2 exchange at two sitesin Arctic tundra ecosystems during the growing season usinga spectral vegetation index, Int. J. Remote Sens., 20, 638–698,1999.

Moffat, A. M., Papale, D., Reichstein, M., Hollinger, D. Y.,Richardson, A. D., Barr, A. G., Beckstein, C., Braswell, B. H.,Churkina, G., Desai, A. R., Falge, E., Gove, J. H., Heimann,M., Hui, D., Jarvis, A. J., Kattge, J., Noormets, A., and Stauch,V. J.: Comprehensive comparison of gap-filling techniques foreddy covariance net carbon fluxes, Agr. Forest Meteorol., 147,209–232, 2007.

Malenovsky, Z., Mishra, K. B., Zemek, F., Rascher, U., Nebal, L.:Scientific and technical challenges in remote sensing of plantcanopy reflectance and fluorescence, J. Exp. Botany, 60, 2987–3000, 2009.

Moya, I., Camenen, L., Evain, S., Goulas, Y., Cerovic, Z. G., La-touche, G., Flexas, J., and Ounis, A.: A new instrument for pas-sive remote sensing 1. Measurements of sunlight-induced chloro-phyll fluorescence, Remote Sens. Environ., 91, 186–197, 2004.

Olofsson, P., Lagergren, F., Lindroth, A., Lindstrm, J., Klemedts-son, L., Kutsch, W., and Eklundh, L.: Towards operational re-mote sensing of forest carbon balance across Northern Europe,Biogeosciences, 5, 817–832, 2008,http://www.biogeosciences.net/5/817/2008/.

Reichstein, M., Falge, E., Baldocchi, D., Papale, D., Aubinet,M., Berbigier, P., Bernhofer, C., Buchmann, N., Gilmanov, T.,Granier, A., Grunwald, T., Havrankova, K., Ilvesniemi, H.,Janous, D., Knohl, A., Laurela, T., Lohila, A., Loustau, D., Mat-teucci, G., Meyers, T., Miglietta, F., Ourcival, J. -M., Pumpanen,J., Rambal, S., Rotenberg, E., Sanz, M., Tenhunen, J., Seufert,G., Vaccari, F., Vesala, T., Yakir, D., and Valentini, R.: Onthe separation of net ecosystem exchange into assimilation andecosystem respiration: review and improved algorithm, GlobalChange Biol., 11, 1424–1439, 2005.

Richardson, A. D. and Hollinger, D. Y.: A method to estimate theadditional uncertainty in gap-filled NEE resulting from long gapsin the CO2 flux record, Agr. Forest Meteorol., 147, 199–208,2007.

Richardson, A. D., Jenkins, J. P., Braswell, B. H., Hollinger, D. Y.,Ollinger, S. V., and Smith, M.-L.: Use of digital webcam imagesto track spring green-up in a deciduous forest, Oecologia, 152,323–334, 2007.

Rocha, A. V. and Shaver, G. R.: Advantages of a two band EVI cal-culated from solar and photosynthetically active radiation fluxes,Agr. For. Meteorol., 149, 1560–1563, 2009.

Rouse, J. W., Haas, R. H., Schell, J. A., and Deering, D. W.: Mon-itoring Vegetation Systems in the Great Plains with ERTS, in:Third ERTS Symposium, NASA SP-353, Vol. 1, 309–317, USGovernment Printing Office, Washington, DC, 1973.

Ruimy, A., Jarvis, P. G., Baldocchi, D. D., and Saugier, B.: CO2fluxes over plant canopies and solar radiation: a review, Adv.Ecol. Res., 26, 1–68, 1995.

Running, S. W., Baldocchi, D. D., Turner, D. P., Gower, S. T.,Bakwin, P. S., and Hibbard, K. A.: A global terrestrial monitor-ing network integrating tower fluxes, flask sampling, ecosystemmodelling and EOS satellite data, Remote Sens. Environ., 70,108–127, 1999.

Sims, D. A., Rahman, A. F., Cordova, V. D., Baldocchi, D. D.,Flanagan, L. B., Goldstein, A. H., Hollinger, D. Y., Misson, L.,Monson, R. K., Schmid, H. P., Wofsy, S. C., and Xu, L.: Middayvalues of gross CO 2 flux and light use efficiency during satelliteoverpasses can be used to directly estimate eight-day mean flux,Agr. Forest Meteorol., 131, 1–12, 2005.

Smith, E. L.: The influence of light and carbon dioxide on photo-synthesis, J. Gen. Physio., 20, 807–830, 1937.

Solomon, S., Quin, D., Manning, M., Chen, Z., Marquis, M., Av-eryt, K. B., Tignor, M., and Miller, H. L.: Climate Change 2007:The Physical Science Basis. Contribution of Working Group I tothe Fourth Assessment Report of the Intergovernmental Panel onClimate Change, Cambridge University Press, Cambridge, 2007.

Steffen, W., Noble, I., Canadell, J., Apps, M., Schulze, E. -D.,Jarvis, P. G., Baldocchi, D., Ciais, P., Cramer, W., Ehleringer,J., Farquhar, G., Field, C. B., Ghazi, A., Gifford, R., Heimann,M., Houghton, R., Kabat, P., Korner, Ch., Lambin, E., Linder, S.,Mooney, H. A., Murdiyarso, D., Post, W. M., Prentice, C., Rau-pach, M. R., Schimel, D. S., Shvidenko, A., and Valentini, R.:The terrestrial carbon cycle: implications for the Kyoto protocol,Science, 280, 1393–1394, 1998.

Turner, D. P., Guzy, M., Lefsky, M. A., Rits, W. D., Van Tuyl,S., and Law, B. E.: Monitoring forest carbon sequestration withremote sensing and carbon cycle modelling, Environ. Manage-ment, 33, 457–466, 2004.

van der Tol, C., Verhoef, W., Timmermans, J., Verhoef, A., and Su,Z.: An integrated model of soil-canopy spectral radiances, pho-tosynthesis, fluorescence, temperature and energy balance, Bio-geosciences, 6, 3109–3129, 2009,http://www.biogeosciences.net/6/3109/2009/.

Wang, Q., Tenhunen, J., Dinh, N. Q., Reichstein, M., Vesala, T., andKeronen, P.: Similarities in ground- and satellite-based NDVItime series and their relationship to physiological activity of aScots pine forest in Finland, Remote Sens. Environ., 93, 225–237, 2004.

Wang, Q., Tenhunen, J., and Vesala, T.: Gross primary produc-tion simulation in a coniferous forest using a daily gas exchangemodel with seasonal change of leaf physiological parameters de-rived from remote sensing data, Int. J. Remote Sens., 30, 3013–3025, 2009.





Wilson, T. B. and Meyers, T. P.: Determining vegetation indicesfrom solar and photosynthetically active radiation fluxes, Agr.Forest Meteorol., 144, 160–179, 2007.

Wohlfahrt, G., Sapinsky, S., Tappeiner, U., and Cernusca, A.: Es-timation of plant area index of grasslands from measurementsof canopy radiation profiles, Agr. Forest Meteorol., 109, 1–12,2001.

Wohlfahrt, G., Hammerle, A., Haslwanter, A., Bahn, M., Tappeiner,U., Cernusca, A.: Seasonal and inter-annual variability of thenet ecosystem CO2 exchange of a temperate mountain grass-land: Effects of weather and management, J. Geophys. Res., 113,D08110, doi:10.1029/2007JD009286, 2008.


Documents

Estimating carbon dioxide fluxes from temperate mountain grasslands using broad-band vegetation indices